X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=filterseqscommand.cpp;h=1b3b3cfd415cd51605fe3c09a8b3b5a9f11b160d;hb=c6e5f84783b39a558119835963dec3c4facb0154;hp=6d894d866a2930eb009d20627cb03f509b94ef2e;hpb=cb9c5129766d46fa1a266f724c1ebd41047f3a03;p=mothur.git diff --git a/filterseqscommand.cpp b/filterseqscommand.cpp index 6d894d8..1b3b3cf 100644 --- a/filterseqscommand.cpp +++ b/filterseqscommand.cpp @@ -28,7 +28,7 @@ FilterSeqsCommand::FilterSeqsCommand(string option) { OptionParser parser(option); map parameters = parser.getParameters(); - ValidParameters validParameter; + ValidParameters validParameter("filter.seqs"); map::iterator it; //check to make sure all parameters are valid for command @@ -44,7 +44,7 @@ FilterSeqsCommand::FilterSeqsCommand(string option) { it = parameters.find("fasta"); //user has given a template file if(it != parameters.end()){ - path = hasPath(it->second); + path = m->hasPath(it->second); //if the user has not given a path then, add inputdir. else leave path alone. if (path == "") { parameters["fasta"] = inputDir + it->second; } } @@ -52,7 +52,7 @@ FilterSeqsCommand::FilterSeqsCommand(string option) { it = parameters.find("hard"); //user has given a template file if(it != parameters.end()){ - path = hasPath(it->second); + path = m->hasPath(it->second); //if the user has not given a path then, add inputdir. else leave path alone. if (path == "") { parameters["hard"] = inputDir + it->second; } } @@ -62,26 +62,37 @@ FilterSeqsCommand::FilterSeqsCommand(string option) { fasta = validParameter.validFile(parameters, "fasta", false); if (fasta == "not found") { m->mothurOut("fasta is a required parameter for the filter.seqs command."); m->mothurOutEndLine(); abort = true; } else { - splitAtDash(fasta, fastafileNames); + m->splitAtDash(fasta, fastafileNames); //go through files and make sure they are good, if not, then disregard them for (int i = 0; i < fastafileNames.size(); i++) { if (inputDir != "") { - string path = hasPath(fastafileNames[i]); + string path = m->hasPath(fastafileNames[i]); //if the user has not given a path then, add inputdir. else leave path alone. if (path == "") { fastafileNames[i] = inputDir + fastafileNames[i]; } } - int ableToOpen; ifstream in; - ableToOpen = openInputFile(fastafileNames[i], in); + int ableToOpen = m->openInputFile(fastafileNames[i], in, "noerror"); + + //if you can't open it, try default location + if (ableToOpen == 1) { + if (m->getDefaultPath() != "") { //default path is set + string tryPath = m->getDefaultPath() + m->getSimpleName(fastafileNames[i]); + m->mothurOut("Unable to open " + fastafileNames[i] + ". Trying default " + tryPath); m->mothurOutEndLine(); + ableToOpen = m->openInputFile(tryPath, in, "noerror"); + fastafileNames[i] = tryPath; + } + } + in.close(); + if (ableToOpen == 1) { - m->mothurOut(fastafileNames[i] + " will be disregarded."); m->mothurOutEndLine(); + m->mothurOut("Unable to open " + fastafileNames[i] + ". It will be disregarded."); m->mothurOutEndLine(); //erase from file list fastafileNames.erase(fastafileNames.begin()+i); i--; }else{ - string simpleName = getSimpleName(fastafileNames[i]); + string simpleName = m->getSimpleName(fastafileNames[i]); filterFileName += simpleName.substr(0, simpleName.find_first_of('.')); } in.close(); @@ -95,13 +106,16 @@ FilterSeqsCommand::FilterSeqsCommand(string option) { //if the user changes the output directory command factory will send this info to us in the output parameter outputDir = validParameter.validFile(parameters, "outputdir", false); if (outputDir == "not found"){ outputDir = ""; - outputDir += hasPath(fastafileNames[0]); //if user entered a file with a path then preserve it + outputDir += m->hasPath(fastafileNames[0]); //if user entered a file with a path then preserve it } } //check for optional parameter and set defaults // ...at some point should added some additional type checking... string temp; + hard = validParameter.validFile(parameters, "hard", true); if (hard == "not found") { hard = ""; } + else if (hard == "not open") { hard = ""; abort = true; } + temp = validParameter.validFile(parameters, "trump", false); if (temp == "not found") { temp = "*"; } trump = temp[0]; @@ -111,13 +125,13 @@ FilterSeqsCommand::FilterSeqsCommand(string option) { temp = validParameter.validFile(parameters, "processors", false); if (temp == "not found"){ temp = "1"; } convert(temp, processors); - hard = validParameter.validFile(parameters, "hard", true); if (hard == "not found") { hard = ""; } - else if (hard == "not open") { abort = true; } - - vertical = validParameter.validFile(parameters, "vertical", false); if (vertical == "not found") { vertical = "T"; } + vertical = validParameter.validFile(parameters, "vertical", false); + if (vertical == "not found") { + if ((hard == "") && (trump == '*') && (soft == 0)) { vertical = "T"; } //you have not given a hard file or set the trump char. + else { vertical = "F"; } + } numSeqs = 0; - } } @@ -138,11 +152,11 @@ void FilterSeqsCommand::help(){ m->mothurOut("For example: fasta=abrecovery.fasta-amazon.fasta \n"); m->mothurOut("The trump parameter .... The default is ...\n"); m->mothurOut("The soft parameter .... The default is ....\n"); - m->mothurOut("The hard parameter .... The default is ....\n"); - m->mothurOut("The vertical parameter .... The default is T.\n"); + m->mothurOut("The hard parameter allows you to enter a file containing the filter you want to use.\n"); + m->mothurOut("The vertical parameter removes columns where all sequences contain a gap character. The default is T.\n"); m->mothurOut("The filter.seqs command should be in the following format: \n"); - m->mothurOut("filter.seqs(fasta=yourFastaFile, trump=yourTrump, soft=yourSoft, hard=yourHard, vertical=yourVertical) \n"); - m->mothurOut("Example filter.seqs(fasta=abrecovery.fasta, trump=..., soft=..., hard=..., vertical=T).\n"); + m->mothurOut("filter.seqs(fasta=yourFastaFile, trump=yourTrump) \n"); + m->mothurOut("Example filter.seqs(fasta=abrecovery.fasta, trump=.).\n"); m->mothurOut("Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFasta).\n\n"); } @@ -160,31 +174,48 @@ int FilterSeqsCommand::execute() { if (abort == true) { return 0; } ifstream inFASTA; - openInputFile(fastafileNames[0], inFASTA); + m->openInputFile(fastafileNames[0], inFASTA); Sequence testSeq(inFASTA); alignmentLength = testSeq.getAlignLength(); inFASTA.close(); ////////////create filter///////////////// + m->mothurOut("Creating Filter... "); m->mothurOutEndLine(); filter = createFilter(); + m->mothurOutEndLine(); m->mothurOutEndLine(); + if (m->control_pressed) { return 0; } + #ifdef USE_MPI + int pid; + MPI_Comm_rank(MPI_COMM_WORLD, &pid); + + if (pid == 0) { //only one process should output the filter + #endif + ofstream outFilter; string filterFile = outputDir + filterFileName + ".filter"; - openOutputFile(filterFile, outFilter); + m->openOutputFile(filterFile, outFilter); outFilter << filter << endl; outFilter.close(); outputNames.push_back(filterFile); + #ifdef USE_MPI + } + #endif ////////////run filter///////////////// + m->mothurOut("Running Filter... "); m->mothurOutEndLine(); + filterSequences(); - + + m->mothurOutEndLine(); m->mothurOutEndLine(); + int filteredLength = 0; for(int i=0;igetRootName(m->getSimpleName(fastafileNames[s])) + "filter.fasta"; #ifdef USE_MPI - int pid, start, end; + int pid, start, end, numSeqsPerProcessor, num; int tag = 2001; + vectorMPIPos; MPI_Status status; MPI_Comm_size(MPI_COMM_WORLD, &processors); //set processors to the number of mpi processes running MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are MPI_File outMPI; + MPI_File tempMPI; MPI_File inMPI; int outMode=MPI_MODE_CREATE|MPI_MODE_WRONLY; int inMode=MPI_MODE_RDONLY; - char outFilename[filteredFasta.length()]; + char outFilename[1024]; strcpy(outFilename, filteredFasta.c_str()); - - char inFileName[fastafileNames[s].length()]; + + char inFileName[1024]; strcpy(inFileName, fastafileNames[s].c_str()); MPI_File_open(MPI_COMM_WORLD, inFileName, inMode, MPI_INFO_NULL, &inMPI); //comm, filename, mode, info, filepointer MPI_File_open(MPI_COMM_WORLD, outFilename, outMode, MPI_INFO_NULL, &outMPI); - + + if (m->control_pressed) { MPI_File_close(&inMPI); MPI_File_close(&outMPI); return 0; } + if (pid == 0) { //you are the root process - setLines(fastafileNames[s]); - - for (int j = 0; j < lines.size(); j++) { //each process - if (j != 0) { //don't send to yourself - MPI_Send(&lines[j]->start, 1, MPI_INT, j, tag, MPI_COMM_WORLD); //start position in file - MPI_Send(&bufferSizes[j], 1, MPI_INT, j, tag, MPI_COMM_WORLD); //how bytes for the read - } + MPIPos = m->setFilePosFasta(fastafileNames[s], num); //fills MPIPos, returns numSeqs + numSeqs += num; + + //send file positions to all processes + for(int i = 1; i < processors; i++) { + MPI_Send(&num, 1, MPI_INT, i, tag, MPI_COMM_WORLD); + MPI_Send(&MPIPos[0], (num+1), MPI_LONG, i, tag, MPI_COMM_WORLD); } - //read your peice of file - char buf[bufferSizes[0]]; - MPI_File_read_at(inMPI, lines[0]->start, buf, bufferSizes[0], MPI_CHAR, &status); - istringstream iss (buf,istringstream::in); + //figure out how many sequences you have to do + numSeqsPerProcessor = num / processors; + int startIndex = pid * numSeqsPerProcessor; + if(pid == (processors - 1)){ numSeqsPerProcessor = num - pid * numSeqsPerProcessor; } + //do your part - driverMPIRun(iss, outMPI); + driverMPIRun(startIndex, numSeqsPerProcessor, inMPI, outMPI, MPIPos); + + if (m->control_pressed) { MPI_File_close(&inMPI); MPI_File_close(&outMPI); return 0; } //wait on chidren for(int i = 1; i < processors; i++) { @@ -270,19 +310,22 @@ int FilterSeqsCommand::filterSequences() { } }else { //you are a child process - //receive your section of file - int startPos, numLines, bufferSize; - MPI_Recv(&startPos, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status); - MPI_Recv(&bufferSize, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status); - - //read your peice of file - char buf2[bufferSize]; - MPI_File_read_at(inMPI, startPos, buf2, bufferSize, MPI_CHAR, &status); - istringstream iss (buf2,istringstream::in); + MPI_Recv(&num, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status); + MPIPos.resize(num+1); + numSeqs += num; + MPI_Recv(&MPIPos[0], (num+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status); + + //figure out how many sequences you have to align + numSeqsPerProcessor = num / processors; + int startIndex = pid * numSeqsPerProcessor; + if(pid == (processors - 1)){ numSeqsPerProcessor = num - pid * numSeqsPerProcessor; } + + + //align your part + driverMPIRun(startIndex, numSeqsPerProcessor, inMPI, outMPI, MPIPos); + + if (m->control_pressed) { MPI_File_close(&inMPI); MPI_File_close(&outMPI); return 0; } - //do your part - driverMPIRun(iss, outMPI); - char buf[4]; strcpy(buf, "done"); @@ -292,41 +335,35 @@ int FilterSeqsCommand::filterSequences() { MPI_File_close(&outMPI); MPI_File_close(&inMPI); + MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case #else + vector positions = m->divideFile(fastafileNames[s], processors); + + for (int i = 0; i < (positions.size()-1); i++) { + lines.push_back(new linePair(positions[i], positions[(i+1)])); + } #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) if(processors == 1){ - ifstream inFASTA; - openInputFile(fastafileNames[s], inFASTA); - int numFastaSeqs=count(istreambuf_iterator(inFASTA),istreambuf_iterator(), '>'); - inFASTA.close(); - - lines.push_back(new linePair(0, numFastaSeqs)); - - driverRunFilter(filter, filteredFasta, fastafileNames[s], lines[0]); + int numFastaSeqs = driverRunFilter(filter, filteredFasta, fastafileNames[s], lines[0]); + numSeqs += numFastaSeqs; }else{ - setLines(fastafileNames[s]); - createProcessesRunFilter(filter, fastafileNames[s]); - + int numFastaSeqs = createProcessesRunFilter(filter, fastafileNames[s]); + numSeqs += numFastaSeqs; + rename((fastafileNames[s] + toString(processIDS[0]) + ".temp").c_str(), filteredFasta.c_str()); //append fasta files for(int i=1;iappendFiles((fastafileNames[s] + toString(processIDS[i]) + ".temp"), filteredFasta); remove((fastafileNames[s] + toString(processIDS[i]) + ".temp").c_str()); } } if (m->control_pressed) { return 1; } #else - ifstream inFASTA; - openInputFile(fastafileNames[s], inFASTA); - int numFastaSeqs=count(istreambuf_iterator(inFASTA),istreambuf_iterator(), '>'); - inFASTA.close(); - - lines.push_back(new linePair(0, numFastaSeqs)); - - driverRunFilter(filter, filteredFasta, fastafileNames[s], lines[0]); + int numFastaSeqs = driverRunFilter(filter, filteredFasta, fastafileNames[s], lines[0]); + numSeqs += numFastaSeqs; if (m->control_pressed) { return 1; } #endif @@ -341,16 +378,29 @@ int FilterSeqsCommand::filterSequences() { exit(1); } } +#ifdef USE_MPI /**************************************************************************************/ -int FilterSeqsCommand::driverMPIRun(istringstream& in, MPI_File& outMPI) { +int FilterSeqsCommand::driverMPIRun(int start, int num, MPI_File& inMPI, MPI_File& outMPI, vector& MPIPos) { try { string outputString = ""; int count = 0; MPI_Status status; - while (!in.eof()) { + for(int i=0;icontrol_pressed) { return 0; } + + //read next sequence + int length = MPIPos[start+i+1] - MPIPos[start+i]; + char* buf4 = new char[length]; + MPI_File_read_at(inMPI, MPIPos[start+i], buf4, length, MPI_CHAR, &status); - Sequence seq(in); gobble(in); + string tempBuf = buf4; + if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); } + istringstream iss (tempBuf,istringstream::in); + delete buf4; + + Sequence seq(iss); m->gobble(iss); if (seq.getName() != "") { string align = seq.getAligned(); @@ -368,26 +418,31 @@ int FilterSeqsCommand::driverMPIRun(istringstream& in, MPI_File& outMPI) { if(count % 10 == 0){ //output to file //send results to parent int length = outputString.length(); - char buf[length]; - strcpy(buf, outputString.c_str()); + char* buf = new char[length]; + memcpy(buf, outputString.c_str(), length); MPI_File_write_shared(outMPI, buf, length, MPI_CHAR, &status); outputString = ""; + delete buf; } } + + if((i+1) % 100 == 0){ cout << (i+1) << endl; m->mothurOutJustToLog(toString(i+1) + "\n"); } } if(outputString != ""){ //output to file //send results to parent int length = outputString.length(); - char buf[length]; - strcpy(buf, outputString.c_str()); + char* buf = new char[length]; + memcpy(buf, outputString.c_str(), length); MPI_File_write_shared(outMPI, buf, length, MPI_CHAR, &status); outputString = ""; + delete buf; } - + + if((num) % 100 != 0){ cout << (num) << endl; m->mothurOutJustToLog(toString(num) + "\n"); } return 0; } @@ -396,22 +451,26 @@ int FilterSeqsCommand::driverMPIRun(istringstream& in, MPI_File& outMPI) { exit(1); } } +#endif /**************************************************************************************/ -int FilterSeqsCommand::driverRunFilter(string F, string outputFilename, string inputFilename, linePair* line) { +int FilterSeqsCommand::driverRunFilter(string F, string outputFilename, string inputFilename, linePair* filePos) { try { ofstream out; - openOutputFile(outputFilename, out); + m->openOutputFile(outputFilename, out); ifstream in; - openInputFile(inputFilename, in); + m->openInputFile(inputFilename, in); - in.seekg(line->start); - - for(int i=0;inumSeqs;i++){ + in.seekg(filePos->start); + + bool done = false; + int count = 0; + + while (!done) { if (m->control_pressed) { in.close(); out.close(); return 0; } - Sequence seq(in); + Sequence seq(in); m->gobble(in); if (seq.getName() != "") { string align = seq.getAligned(); string filterSeq = ""; @@ -423,13 +482,23 @@ int FilterSeqsCommand::driverRunFilter(string F, string outputFilename, string i } out << '>' << seq.getName() << endl << filterSeq << endl; - } - gobble(in); + count++; + } + + unsigned long int pos = in.tellg(); + if ((pos == -1) || (pos >= filePos->end)) { break; } + + //report progress + if((count) % 100 == 0){ m->mothurOut(toString(count)); m->mothurOutEndLine(); } } + //report progress + if((count) % 100 != 0){ m->mothurOut(toString(count)); m->mothurOutEndLine(); } + + out.close(); in.close(); - return 0; + return count; } catch(exception& e) { m->errorOut(e, "FilterSeqsCommand", "driverRunFilter"); @@ -442,7 +511,7 @@ int FilterSeqsCommand::createProcessesRunFilter(string F, string filename) { try { #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) int process = 0; - int exitCommand = 1; + int num = 0; processIDS.clear(); //loop through and create all the processes you want @@ -454,7 +523,15 @@ int FilterSeqsCommand::createProcessesRunFilter(string F, string filename) { process++; }else if (pid == 0){ string filteredFasta = filename + toString(getpid()) + ".temp"; - driverRunFilter(F, filteredFasta, filename, lines[process]); + num = driverRunFilter(F, filteredFasta, filename, lines[process]); + + //pass numSeqs to parent + ofstream out; + string tempFile = filename + toString(getpid()) + ".num.temp"; + m->openOutputFile(tempFile, out); + out << num << endl; + out.close(); + exit(0); }else { m->mothurOut("unable to spawn the necessary processes."); m->mothurOutEndLine(); exit(0); } } @@ -463,9 +540,18 @@ int FilterSeqsCommand::createProcessesRunFilter(string F, string filename) { for (int i=0;iopenInputFile(tempFile, in); + if (!in.eof()) { int tempNum = 0; in >> tempNum; num += tempNum; } + in.close(); remove(tempFile.c_str()); } + - return exitCommand; + return num; #endif } catch(exception& e) { @@ -476,8 +562,7 @@ int FilterSeqsCommand::createProcessesRunFilter(string F, string filename) { /**************************************************************************************/ string FilterSeqsCommand::createFilter() { try { - string filterString = ""; - + string filterString = ""; Filters F; if (soft != 0) { F.setSoft(soft); } @@ -485,7 +570,7 @@ string FilterSeqsCommand::createFilter() { F.setLength(alignmentLength); - if(soft != 0 || isTrue(vertical)){ + if(trump != '*' || m->isTrue(vertical) || soft != 0){ F.initialize(); } @@ -493,131 +578,92 @@ string FilterSeqsCommand::createFilter() { else { F.setFilter(string(alignmentLength, '1')); } numSeqs = 0; - - if(trump != '*' || isTrue(vertical) || soft != 0){ + if(trump != '*' || m->isTrue(vertical) || soft != 0){ for (int s = 0; s < fastafileNames.size(); s++) { + for (int i = 0; i < lines.size(); i++) { delete lines[i]; } lines.clear(); + #ifdef USE_MPI - int pid, rc, ierr; - int Atag = 1; int Ttag = 2; int Ctag = 3; int Gtag = 4; int Gaptag = 5; + int pid, numSeqsPerProcessor, num; int tag = 2001; + vector MPIPos; MPI_Status status; - MPI_File in; - rc = MPI_Comm_size(MPI_COMM_WORLD, &processors); - rc = MPI_Comm_rank(MPI_COMM_WORLD, &pid); + MPI_File inMPI; + MPI_Comm_size(MPI_COMM_WORLD, &processors); + MPI_Comm_rank(MPI_COMM_WORLD, &pid); - char tempFileName[fastafileNames[s].length()]; + //char* tempFileName = new char(fastafileNames[s].length()); + //tempFileName = &(fastafileNames[s][0]); + + char tempFileName[1024]; strcpy(tempFileName, fastafileNames[s].c_str()); - cout << pid << " tempFileName " << tempFileName << endl; - MPI_File_open(MPI_COMM_WORLD, tempFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &in); //comm, filename, mode, info, filepointer - cout << pid << " here" << endl; + + MPI_File_open(MPI_COMM_WORLD, tempFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI); //comm, filename, mode, info, filepointer + + if (m->control_pressed) { MPI_File_close(&inMPI); return 0; } + if (pid == 0) { //you are the root process - setLines(fastafileNames[s]); - cout << pid << " after setlines" << endl; - for (int j = 0; j < lines.size(); j++) { //each process - if (j != 0) { //don't send to yourself - MPI_Send(&lines[j]->start, 1, MPI_INT, j, tag, MPI_COMM_WORLD); //start position in file - MPI_Send(&bufferSizes[j], 1, MPI_INT, j, tag, MPI_COMM_WORLD); //how bytes for the read - } + MPIPos = m->setFilePosFasta(fastafileNames[s], num); //fills MPIPos, returns numSeqs + numSeqs += num; + + //send file positions to all processes + for(int i = 1; i < processors; i++) { + MPI_Send(&num, 1, MPI_INT, i, tag, MPI_COMM_WORLD); + MPI_Send(&MPIPos[0], (num+1), MPI_LONG, i, tag, MPI_COMM_WORLD); } - cout << pid << " done sending" << endl; - char buf[bufferSizes[0]]; - MPI_File_read_at(in, 0, buf, bufferSizes[0], MPI_CHAR, &status); - cout << pid << " done reading" << endl; - string tempBuf = buf; - if (tempBuf.length() > bufferSizes[0]) { tempBuf = tempBuf.substr(0, bufferSizes[0]); } - - MPICreateFilter(F, tempBuf); - cout << pid << "done with mpi create filter " << endl; - if (m->control_pressed) { MPI_File_close(&in); return filterString; } - - vector temp; temp.resize(alignmentLength+1); + + //figure out how many sequences you have to do + numSeqsPerProcessor = num / processors; + int startIndex = pid * numSeqsPerProcessor; + if(pid == (processors - 1)){ numSeqsPerProcessor = num - pid * numSeqsPerProcessor; } - //get the frequencies from the child processes - for(int i = 0; i < ((processors-1)*5); i++) { - MPI_Recv(&temp[0], (alignmentLength+1), MPI_INT, MPI_ANY_SOURCE, tag, MPI_COMM_WORLD, &status); - int receiveTag = temp[temp.size()-1]; //child process added a int to the end to indicate what letter count this is for - - if (receiveTag == Atag) { //you are recieveing the A frequencies - for (int k = 0; k < alignmentLength; k++) { F.a[k] += temp[k]; } - }else if (receiveTag == Ttag) { //you are recieveing the T frequencies - for (int k = 0; k < alignmentLength; k++) { F.t[k] += temp[k]; } - }else if (receiveTag == Ctag) { //you are recieveing the C frequencies - for (int k = 0; k < alignmentLength; k++) { F.c[k] += temp[k]; } - }else if (receiveTag == Gtag) { //you are recieveing the G frequencies - for (int k = 0; k < alignmentLength; k++) { F.g[k] += temp[k]; } - }else if (receiveTag == Gaptag) { //you are recieveing the gap frequencies - for (int k = 0; k < alignmentLength; k++) { F.gap[k] += temp[k]; } - } - } - + + //do your part + MPICreateFilter(startIndex, numSeqsPerProcessor, F, inMPI, MPIPos); + if (m->control_pressed) { MPI_File_close(&inMPI); return 0; } + }else { //i am the child process - cout << pid << endl; - int startPos, bufferSize; - ierr = MPI_Recv(&startPos, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status); - ierr = MPI_Recv(&bufferSize, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status); - cout << pid << '\t' << startPos << '\t' << bufferSize << endl; - //send freqs - char buf2[bufferSize]; - MPI_File_read_at( in, startPos, buf2, bufferSize, MPI_CHAR, &status); - - string tempBuf = buf2; - if (tempBuf.length() > bufferSize) { tempBuf = tempBuf.substr(0, bufferSize); } - - MPICreateFilter(F, tempBuf); - cout << pid << "done with mpi create filter " << endl; - if (m->control_pressed) { MPI_File_close(&in); return filterString; } + MPI_Recv(&num, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status); + MPIPos.resize(num+1); + numSeqs += num; + MPI_Recv(&MPIPos[0], (num+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status); + + //figure out how many sequences you have to align + numSeqsPerProcessor = num / processors; + int startIndex = pid * numSeqsPerProcessor; + if(pid == (processors - 1)){ numSeqsPerProcessor = num - pid * numSeqsPerProcessor; } + + + //do your part + MPICreateFilter(startIndex, numSeqsPerProcessor, F, inMPI, MPIPos); - //send my fequency counts - F.a.push_back(Atag); - int ierr = MPI_Send(&(F.a[0]), (alignmentLength+1), MPI_INT, 0, tag, MPI_COMM_WORLD); - F.t.push_back(Ttag); - ierr = MPI_Send (&(F.t[0]), (alignmentLength+1), MPI_INT, 0, tag, MPI_COMM_WORLD); - F.c.push_back(Ctag); - ierr = MPI_Send(&(F.c[0]), (alignmentLength+1), MPI_INT, 0, tag, MPI_COMM_WORLD); - F.g.push_back(Gtag); - ierr = MPI_Send(&(F.g[0]), (alignmentLength+1), MPI_INT, 0, tag, MPI_COMM_WORLD); - F.gap.push_back(Gaptag); - ierr = MPI_Send(&(F.gap[0]), (alignmentLength+1), MPI_INT, 0, tag, MPI_COMM_WORLD); + if (m->control_pressed) { MPI_File_close(&inMPI); return 0; } } - MPI_Barrier(MPI_COMM_WORLD); - MPI_File_close(&in); + MPI_File_close(&inMPI); + MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case #else + vector positions = m->divideFile(fastafileNames[s], processors); + + for (int i = 0; i < (positions.size()-1); i++) { + lines.push_back(new linePair(positions[i], positions[(i+1)])); + } #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) if(processors == 1){ - ifstream inFASTA; - openInputFile(fastafileNames[s], inFASTA); - int numFastaSeqs=count(istreambuf_iterator(inFASTA),istreambuf_iterator(), '>'); - inFASTA.close(); - + int numFastaSeqs = driverCreateFilter(F, fastafileNames[s], lines[0]); numSeqs += numFastaSeqs; - - for (int i = 0; i < lines.size(); i++) { delete lines[i]; } lines.clear(); - lines.push_back(new linePair(0, numFastaSeqs)); - - driverCreateFilter(F, fastafileNames[s], lines[0]); }else{ - setLines(fastafileNames[s]); - createProcessesCreateFilter(F, fastafileNames[s]); + int numFastaSeqs = createProcessesCreateFilter(F, fastafileNames[s]); + numSeqs += numFastaSeqs; } if (m->control_pressed) { return filterString; } #else - ifstream inFASTA; - openInputFile(fastafileNames[s], inFASTA); - int numFastaSeqs=count(istreambuf_iterator(inFASTA),istreambuf_iterator(), '>'); - inFASTA.close(); - + int numFastaSeqs = driverCreateFilter(F, fastafileNames[s], lines[0]); numSeqs += numFastaSeqs; - - for (int i = 0; i < lines.size(); i++) { delete lines[i]; } lines.clear(); - lines.push_back(new linePair(0, numFastaSeqs)); - - driverCreateFilter(F, fastafileNames[s], lines[0]); if (m->control_pressed) { return filterString; } #endif #endif @@ -625,13 +671,88 @@ string FilterSeqsCommand::createFilter() { } } - F.setNumSeqs(numSeqs); + +#ifdef USE_MPI + int pid; + int Atag = 1; int Ttag = 2; int Ctag = 3; int Gtag = 4; int Gaptag = 5; + MPI_Status status; + + MPI_Comm_rank(MPI_COMM_WORLD, &pid); + + if(trump != '*' || m->isTrue(vertical) || soft != 0){ + + if (pid == 0) { //only one process should output the filter + + vector temp; temp.resize(alignmentLength+1); + + //get the frequencies from the child processes + for(int i = 1; i < processors; i++) { - if(isTrue(vertical) == 1) { F.doVertical(); } - if(soft != 0) { F.doSoft(); } + for (int j = 0; j < 5; j++) { + + MPI_Recv(&temp[0], (alignmentLength+1), MPI_INT, i, 2001, MPI_COMM_WORLD, &status); + int receiveTag = temp[temp.size()-1]; //child process added a int to the end to indicate what letter count this is for + + if (receiveTag == Atag) { //you are recieveing the A frequencies + for (int k = 0; k < alignmentLength; k++) { F.a[k] += temp[k]; } + }else if (receiveTag == Ttag) { //you are recieveing the T frequencies + for (int k = 0; k < alignmentLength; k++) { F.t[k] += temp[k]; } + }else if (receiveTag == Ctag) { //you are recieveing the C frequencies + for (int k = 0; k < alignmentLength; k++) { F.c[k] += temp[k]; } + }else if (receiveTag == Gtag) { //you are recieveing the G frequencies + for (int k = 0; k < alignmentLength; k++) { F.g[k] += temp[k]; } + }else if (receiveTag == Gaptag) { //you are recieveing the gap frequencies + for (int k = 0; k < alignmentLength; k++) { F.gap[k] += temp[k]; } + } + } + } + }else{ + + //send my fequency counts + F.a.push_back(Atag); + int ierr = MPI_Send(&(F.a[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD); + F.t.push_back(Ttag); + ierr = MPI_Send (&(F.t[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD); + F.c.push_back(Ctag); + ierr = MPI_Send(&(F.c[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD); + F.g.push_back(Gtag); + ierr = MPI_Send(&(F.g[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD); + F.gap.push_back(Gaptag); + ierr = MPI_Send(&(F.gap[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD); + } - filterString = F.getFilter(); + } + + MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case + + if (pid == 0) { //only one process should output the filter +#endif + F.setNumSeqs(numSeqs); + if(m->isTrue(vertical) == 1) { F.doVertical(); } + if(soft != 0) { F.doSoft(); } + filterString = F.getFilter(); + +#ifdef USE_MPI + //send filter string to kids + //for(int i = 1; i < processors; i++) { + // MPI_Send(&filterString[0], alignmentLength, MPI_CHAR, i, 2001, MPI_COMM_WORLD); + //} + MPI_Bcast(&filterString[0], alignmentLength, MPI_CHAR, 0, MPI_COMM_WORLD); + }else{ + //recieve filterString + char* tempBuf = new char[alignmentLength]; + //MPI_Recv(&tempBuf[0], alignmentLength, MPI_CHAR, 0, 2001, MPI_COMM_WORLD, &status); + MPI_Bcast(tempBuf, alignmentLength, MPI_CHAR, 0, MPI_COMM_WORLD); + + filterString = tempBuf; + if (filterString.length() > alignmentLength) { filterString = filterString.substr(0, alignmentLength); } + delete tempBuf; + } + + MPI_Barrier(MPI_COMM_WORLD); +#endif + return filterString; } catch(exception& e) { @@ -640,68 +761,86 @@ string FilterSeqsCommand::createFilter() { } } /**************************************************************************************/ -int FilterSeqsCommand::driverCreateFilter(Filters& F, string filename, linePair* line) { +int FilterSeqsCommand::driverCreateFilter(Filters& F, string filename, linePair* filePos) { try { ifstream in; - openInputFile(filename, in); + m->openInputFile(filename, in); - in.seekg(line->start); - - for(int i=0;inumSeqs;i++){ + in.seekg(filePos->start); + + bool done = false; + int count = 0; + + while (!done) { if (m->control_pressed) { in.close(); return 1; } - Sequence seq(in); + Sequence seq(in); m->gobble(in); if (seq.getName() != "") { if (seq.getAligned().length() != alignmentLength) { m->mothurOut("Sequences are not all the same length, please correct."); m->mothurOutEndLine(); m->control_pressed = true; } - if(trump != '*'){ F.doTrump(seq); } - if(isTrue(vertical) || soft != 0){ F.getFreqs(seq); } + if(trump != '*') { F.doTrump(seq); } + if(m->isTrue(vertical) || soft != 0) { F.getFreqs(seq); } cout.flush(); + count++; } + unsigned long int pos = in.tellg(); + if ((pos == -1) || (pos >= filePos->end)) { break; } + //report progress - if((i+1) % 100 == 0){ m->mothurOut(toString(i+1)); m->mothurOutEndLine(); } + if((count) % 100 == 0){ m->mothurOut(toString(count)); m->mothurOutEndLine(); } } - //report progress - if((line->numSeqs) % 100 != 0){ m->mothurOut(toString(line->numSeqs)); m->mothurOutEndLine(); } - + if((count) % 100 != 0){ m->mothurOut(toString(count)); m->mothurOutEndLine(); } in.close(); - return 0; + return count; } catch(exception& e) { m->errorOut(e, "FilterSeqsCommand", "driverCreateFilter"); exit(1); } } +#ifdef USE_MPI /**************************************************************************************/ -int FilterSeqsCommand::MPICreateFilter(Filters& F, string input) { +int FilterSeqsCommand::MPICreateFilter(int start, int num, Filters& F, MPI_File& inMPI, vector& MPIPos) { try { - vector seqStrings; - parseBuffer(input, seqStrings); + MPI_Status status; + int pid; + MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are - for(int i=0;icontrol_pressed = true; } - - if (m->control_pressed) { return 1; } + if (m->control_pressed) { return 0; } + + //read next sequence + int length = MPIPos[start+i+1] - MPIPos[start+i]; + + char* buf4 = new char[length]; + MPI_File_read_at(inMPI, MPIPos[start+i], buf4, length, MPI_CHAR, &status); - Sequence seq("", seqStrings[i]); + string tempBuf = buf4; + if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); } + istringstream iss (tempBuf,istringstream::in); + delete buf4; + + Sequence seq(iss); + + if (seq.getAligned().length() != alignmentLength) { cout << "Alignment length is " << alignmentLength << " and sequence " << seq.getName() << " has length " << seq.getAligned().length() << ", please correct." << endl; exit(1); } if(trump != '*'){ F.doTrump(seq); } - if(isTrue(vertical) || soft != 0){ F.getFreqs(seq); } + if(m->isTrue(vertical) || soft != 0){ F.getFreqs(seq); } cout.flush(); //report progress - if((i+1) % 100 == 0){ m->mothurOut(toString(i+1)); m->mothurOutEndLine(); } + if((i+1) % 100 == 0){ cout << (i+1) << endl; m->mothurOutJustToLog(toString(i+1) + "\n"); } } //report progress - if((seqStrings.size()) % 100 != 0){ m->mothurOut(toString(seqStrings.size())); m->mothurOutEndLine(); } + if((num) % 100 != 0){ cout << num << endl; m->mothurOutJustToLog(toString(num) + "\n"); } return 0; } @@ -710,25 +849,49 @@ int FilterSeqsCommand::MPICreateFilter(Filters& F, string input) { exit(1); } } - +#endif /**************************************************************************************************/ int FilterSeqsCommand::createProcessesCreateFilter(Filters& F, string filename) { try { #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) int process = 0; - int exitCommand = 1; + int num = 0; processIDS.clear(); //loop through and create all the processes you want while (process != processors) { - int pid = vfork(); + int pid = fork(); if (pid > 0) { processIDS.push_back(pid); //create map from line number to pid so you can append files in correct order later process++; }else if (pid == 0){ - driverCreateFilter(F, filename, lines[process]); + //reset child's filter counts to 0; + F.a.clear(); F.a.resize(alignmentLength, 0); + F.t.clear(); F.t.resize(alignmentLength, 0); + F.g.clear(); F.g.resize(alignmentLength, 0); + F.c.clear(); F.c.resize(alignmentLength, 0); + F.gap.clear(); F.gap.resize(alignmentLength, 0); + + num = driverCreateFilter(F, filename, lines[process]); + + //write out filter counts to file + filename += toString(getpid()) + "filterValues.temp"; + ofstream out; + m->openOutputFile(filename, out); + + out << num << endl; + out << F.getFilter() << endl; + for (int k = 0; k < alignmentLength; k++) { out << F.a[k] << '\t'; } out << endl; + for (int k = 0; k < alignmentLength; k++) { out << F.t[k] << '\t'; } out << endl; + for (int k = 0; k < alignmentLength; k++) { out << F.g[k] << '\t'; } out << endl; + for (int k = 0; k < alignmentLength; k++) { out << F.c[k] << '\t'; } out << endl; + for (int k = 0; k < alignmentLength; k++) { out << F.gap[k] << '\t'; } out << endl; + + //cout << F.getFilter() << endl; + out.close(); + exit(0); }else { m->mothurOut("unable to spawn the necessary processes."); m->mothurOutEndLine(); exit(0); } } @@ -739,94 +902,35 @@ int FilterSeqsCommand::createProcessesCreateFilter(Filters& F, string filename) wait(&temp); } - return exitCommand; -#endif - } - catch(exception& e) { - m->errorOut(e, "FilterSeqsCommand", "createProcessesCreateFilter"); - exit(1); - } -} -/**************************************************************************************************/ - -int FilterSeqsCommand::setLines(string filename) { - try { - for (int i = 0; i < lines.size(); i++) { delete lines[i]; } lines.clear(); - vector positions; - bufferSizes.clear(); - - ifstream inFASTA; - openInputFile(filename, inFASTA); + //parent reads in and combine Filter info + for (int i = 0; i < processIDS.size(); i++) { + string tempFilename = filename + toString(processIDS[i]) + "filterValues.temp"; + ifstream in; + m->openInputFile(tempFilename, in); - string input; - while(!inFASTA.eof()){ - input = getline(inFASTA); + int temp, tempNum; + string tempFilterString; - if (input.length() != 0) { - if(input[0] == '>'){ long int pos = inFASTA.tellg(); positions.push_back(pos - input.length() - 1); } - } - } - inFASTA.close(); - - int numFastaSeqs = positions.size(); - - FILE * pFile; - long size; - - //get num bytes in file - pFile = fopen (filename.c_str(),"rb"); - if (pFile==NULL) perror ("Error opening file"); - else{ - fseek (pFile, 0, SEEK_END); - size=ftell (pFile); - fclose (pFile); - } - - numSeqs += numFastaSeqs; - - int numSeqsPerProcessor = numFastaSeqs / processors; - - for (int i = 0; i < processors; i++) { + in >> tempNum; m->gobble(in); num += tempNum; - long int startPos = positions[ i * numSeqsPerProcessor ]; - if(i == processors - 1){ - numSeqsPerProcessor = numFastaSeqs - i * numSeqsPerProcessor; - bufferSizes.push_back(size - startPos); - }else{ - long int myEnd = positions[ (i+1) * numSeqsPerProcessor ]; - bufferSizes.push_back(myEnd-startPos); - } - lines.push_back(new linePair(startPos, numSeqsPerProcessor)); - } - - return 0; - } - catch(exception& e) { - m->errorOut(e, "FilterSeqsCommand", "setLines"); - exit(1); - } -} -/**************************************************************************************************/ -int FilterSeqsCommand::parseBuffer(string file, vector& seqs) { - try { - istringstream iss (file); //,istringstream::in - string name, seqstring; + in >> tempFilterString; + F.mergeFilter(tempFilterString); - while (!iss.eof()) { - - if (m->control_pressed) { return 0; } - - Sequence seq(iss); gobble(iss); - - if (seq.getName() != "") { - seqs.push_back(seq.getAligned()); - } + for (int k = 0; k < alignmentLength; k++) { in >> temp; F.a[k] += temp; } m->gobble(in); + for (int k = 0; k < alignmentLength; k++) { in >> temp; F.t[k] += temp; } m->gobble(in); + for (int k = 0; k < alignmentLength; k++) { in >> temp; F.g[k] += temp; } m->gobble(in); + for (int k = 0; k < alignmentLength; k++) { in >> temp; F.c[k] += temp; } m->gobble(in); + for (int k = 0; k < alignmentLength; k++) { in >> temp; F.gap[k] += temp; } m->gobble(in); + + in.close(); + remove(tempFilename.c_str()); } - return 0; + return num; +#endif } catch(exception& e) { - m->errorOut(e, "FilterSeqsCommand", "parseBuffer"); + m->errorOut(e, "FilterSeqsCommand", "createProcessesCreateFilter"); exit(1); } }